U.S. Department of Justice Office of Justice Programs Bureau of Justice Statistics special Report Effects of NIBRS on Crime Statistics July 2000, NCJ 178890 -------------------------------------------------------- This file is text only without graphics and many of the tables. A zip archive of the tables in this report in spreadsheet format (.wk1) and the full report including tables and graphics in .pdf format are available from: http://www.ojp.usdoj.gov/bjs/abstract/enibrsc.htm -------------------------------------------------------- By Ramona R. Rantala BJS Statistician with technical support from Thomas J. Edwards FBI Systems Analyst -------------------------------------------------------- Highlights Differences between Summary UCR and NIBRS Summary UCR * Consists of monthly aggregate crime counts for eight Index crimes * Records one offense per incident as determined by hierarchy rule * Hierarchy rule suppresses counts of lesser offenses in multiple-offense incident * Does not distinguish between attempted and completed crimes * Applies hotel rule to burglary * Records rape of females only * Collects weapon information for murder, robbery, and aggravated assault * Provides counts on arrests for the 8 Index crimes and 21 other offenses. NIBRS * Consists of individual incident records for the 8 Index crimes and 38 other offenses with details on -- * Offense * Victim * Offender * Property * Records each offense occurring in incident * Distinguishes between attempted and completed crimes * Expands burglary hotel rule to include rental storage facilities * Records rape of males and females * Restructures definition of assault * Collects weapon information for all violent offenses * Provides details on arrests for the 8 Index crimes and 49 other offenses. Sources of data discrepancies between Summary UCR and NIBRS Human error Misclassifications of -- * Aggravated assault * Motor vehicle theft * Larceny Hierarchy * Suppression of multiple offenses by hierarchy rule Operational * Modified offense definitions * Converting State penal code to FBI crime definitions * Over-reporting because of State practices * Recording every possible offense with which offender may be charged Computer * Inadequate programming --------------------------- When comparing data from the same year for the jurisdictions in this study, NIBRS rates differ only slightly from Summary UCR -- * Murder rates are the same * Rape, robbery, and aggravated assault rates in NIBRS are about 1% higher, on average, than in Summary UCR * NIBRS burglary rates are lower by an average 0.5% * NIBRS larceny rates are higher by an average 3.4% * NIBRS motor vehicle theft rates are higher by an average 4.5%. ---------------------------------------------------------- In 1929 the FBI began the Summary Uniform Crime Reporting (UCR) system, a voluntary program to collect aggregate counts of Index crimes -- murder, rape, robbery, aggravated assault, burglary, motor vehicle theft, larceny, and, after 1978, arson. " . . . [T]he Summary system of the UCR Program has been very serviceable since its inception in 1929, " the FBI notes, "but it pales next to the capabilities and potential of the National Incident -Based Reporting System (NIBRS).***Footnote"1: FBI, CJIS Newsletter, NIBRS edition, 4, 1, p 1.*** "In developing NIBRS, UCR Program managers have provided law enforcement agencies with a standardized, electronic blueprint for storing the NIBRS data within their individual records management systems.... The goals of NIBRS are to enhance the quantity, quality, and timeliness of crime data collection ...and to improve the thodology used in compiling, analyzing, auditing, and publishing the collected crime statistics.***Footnote"2: CJIS Newsletter, p.2.*** As jurisdictions switch from the Summary UCR system to NIBRS, many have expressed concern about the effects incident-based reporting will have on crime statistics. Under Summary UCR agencies report only the most serious Index offense per incident of crime. Because of the new type of crime classification procedures in NIBRS, in a multiple-offense incident more than one offense will be reported. Jurisdictions are concerned about the effect this more thorough reporting will have on their crime statistics. Comparing crime rates from Summary UCR and NIBRS Research for this report measured differences between estimats of the aggravated assault rate, comparing NIBRS to summary format, and identified possible causes of those differences. Differences in the rates for six other Index crime categories are examined to compare with aggravated assault. Arson, which the FBI reports in the "Modified Index" but not the Index crime figures, is not included. The Bureau of Justice Statistics (BJS) selected jurisdictions with full-year NIBRS reporting and nonzero populations (that is, including agencies with a specific population, such as a local police department or sheriff's office, and generally excluding State or county police). In 1991, 269 agencies, covering a population of 4.1 million, met the criteria. By 1996, full-year NIBRS participation had increased to 1,082 agencies, covering a population of 14.8 million. The data came from nine NIBRS- certified States (Idaho, Iowa, Massachusetts, Michigan, North Dakota, South Carolina, Utah, Vermont, and Virginia). Each agency is represented for each year in which it met the selection criteria. Although 156 agencies had a single year of data, the rest of the agencies met the criteria for more than 1 year. The term "case" in this report refers to an "agency-year." A case includes both NIBRS and Summary UCR aggregate crime counts within each agency for each year an agency reported 12 months of NIBRS data. A total of 4,068 cases, comprising 1,131 unique agencies, met these selection criteria. Two FBI datasets were used. In the first dataset, the FBI provided NIBRS offense counts aggregated by agency from 1991 to 1996. Ten crime categories, as defined in the Summary UCR program, were included. Through a computer program the FBI applied the hierarchy rule and converted the NIBRS counts to Summary UCR counts. This process resulted in two crime counts, NIBRS and Summary UCR, for each crime category, agency, and year. The second data set was the 1996 NIBRS incident file. This file provided detail at the incident level, such as which crimes occurred in combination for multiple offense incidents and the number of victims and offenders. These details help explain some of the crime rate differences. Overall results When the crime rates were calculated from NIBRS and Summary UCR data, the average difference between estimates was small. On average the NIBRS Index crime rate was 2% higher. The violent crime rate was higher by less than 1%, and the property crime rate was higher by slightly more than 2%, on average. ---------------------------------- Table 2. Offense categories for Summary UCR and NIBR Summary UCR Offenses and arrests are reported for the following, listed in hierarchical order: Part I (Index) offenses Murder Forcible rape Robbery Aggravated assault Burglary C breaking or entering Motor vehicle theft Larceny Arson (not subject to the hierarchy rule) Arrests only are reported for the following: Part II offenses Curfew and loitering law violations Disorderly conduct Driving under the influence Drug abuse violations Drunkenness Embezzlement Forgery and counterfeiting Fraud Gambling Liquor laws Offenses against family and children Other assaults Prostitution and commercial vice Runaways Sex offenses (except forcible rape and prostitution) Stolen property: buying, receiving, possess Suspicion Vagrancy Vandalism Weapons: carrying, possessing, other All other offenses (except traffic) NIBRS Offenses and arrests are reported for the following, for which a hierarchy does not apply: Group A offenses Arson Assault offenses Bribery Burglary/breaking and entering Counterfeiting/forgery Destruction/damage/vandalism of property Drug/narcotic offenses Embezzlement Extortion/blackmail Fraud offenses Gambling offenses Homicide offenses Kidnaping/abduction Larceny/theft offenses Motor vehicle theft Pornography/obscene material Prostitution offenses Robbery Sex offenses, forcible Sex offenses, nonforcible Stolen property offenses Weapon law violations Arrests only are reported for the following: Group B offenses Bad checks Curfew/loitering/vagrancy Disorderly conduct Driving under the influence Drunkenness Nonviolent family offenses Peeping Tom Runaways Trespassing All other offenses ------------------------------ For murder, rates from NIBRS and Summary UCR are the same because both systems count each homicide and manslaughter victim and exclude justifiable homicide. Rape, robbery, and aggravated assault rates each had an average difference of less than 1% between NIBRS and Summary UCR statistics. NIBRS burglary rates were lower an average of 0.5%. Burglary is the only individual crime category in which the NIBRS rate can be lower than Summary UCR. (See page 5.) Larceny rates were higher by slightly more than 3% and motor vehicle theft rates by 4.5%, on average. Individually, most cases studied showed only slight differences between NIBRS and Summary UCR estimates. For murder, 100% of the cases had identical murder rates from the two counting methods. NIBRS crime rates differ from Summary UCR by 5.5% or less Number Percent of of cases* all cases All Index offenses 3,661 90% Violent offenses 3,937 97 Murder 4,068 100 Rape 3,957 97 Robbery 3,954 97 Assault 3,970 98 Property offenses 3,643 90 Burglary 3,953 97 Motor vehicle theft 3,234 80 Larceny 3,569 88 *Cases are "agency-years" C agencies that submitted 12 months of NIBRS data for a year. In 3,970 cases of aggravated assault, 98% of all cases, the difference between the rates from NIBRS and Summary UCR was 5.5% or less. Results for rape, robbery and total violent crime were similar, with 97% of cases showing differences of 5.5% or less The results for property crime varied. The proportion of cases having a rate change of 5.5% or less was 97% for burglary, 80% for motor vehicle theft, and 88% for larceny. Overall, 90% of cases showed rate changes of 5.5% or less for both total property and total index crime. Jurisdictions with little crime tend to show exaggerated changes in crime rates. For example, a jurisdiction that experienced in a year two robberies, one of which was in conjunction with a murder, would count one robbery under Summary UCR and two under NIBRS. Though the actual count differed by one, the percent difference was 100%. Such jurisdictions should rely on the actual numbers rather than percent differences to express changes in measurement or over time. Differences in data collection and crime classification between NIBRS and Summary UCR Data collection In Summary UCR, for each of the seven Index crime classifications monthly totals are reported on five elements: offenses reported or known to police (including "unfounded" offenses and attempts); unfounded, that is, false or baseless complaints; number of actual offenses (including attempts); total offenses cleared by arrest or exceptional means; and number of clearances involving only persons under age 18. UCR collects weapon information for murder, robbery, and aggravated assault. Attempts are distinguished from completed crimes only for forcible rape. Murder attempts are reported as aggravated assault for both Summary UCR and NIBRS. Both systems also count each victim of violent offenses. For example, if an offender with a weapon simultaneously injures two victims in an assault, both systems would record two aggravated assaults. In incidents with multiple offenses, Summary UCR records only the most serious offense, as determined by the hierarchy rule. Reported separately, arson is not subject to the hierarchy rule and is not included in this study. Collected information on stolen property provides the number of offenses and the monetary value of property stolen for each Index crime. These totals are recorded by where a robbery took place; whether a burglary occurred at a residence or nonresidence, night or day; nature of larcenies; recovery of motor vehicles. Summary UCR collects monetary value of stolen and recovered property by type of property on a supplemental form. Arrest data are collected for 21 Part II offenses as well as the 8 Index crimes. Instead of collecting aggregate counts by offense category, NIBRS collects detailed information by crime incident, including offense, victim, offender, property and arrest information. Each incident is assigned a number for identifying all information that pertains to it. Details are collected for each occurrence within 8 Index and 38 other specific Group A offenses in NIBRS. (For a list of specific Part I, Part II, Group A, and Group B offenses, These details include information on weapons, location, time of day, alcohol or drug use by offender and hate/bias motivation. Attempts are distinguished from completed crimes for all offenses. In incidents with multiple offenses, all offenses are reported. Data collected on each victim, offend-er, and arrestee include age, sex, and race. Relationship of victim to offender is noted. NIBRS also collects detailed information on type, value, description and recovery of property; type and quantity of drugs; and number of stolen and recovered motor vehicles. Arrest data are collected for 11 Group B offenses in addition to the 22 Group A offense categories. Crime classification Summary UCR and NIBRS differ in how they classify or define several types of crimes. Justifiable homicide is first recorded as murder in Summary UCR, then as unfounded. NIBRS records justifiable homicide separately, not including it with murder and manslaughter. Neither includes justifiable homicide in total counts, so that homicide statistics are not affected. Other differences affect numbers of specific offenses. NIBRS includes male victims of female offenders in forcible rape, which would account for part of the higher rate that was observed. ---------------------------------- Table 3. Information that NIBRS records on each crime incident Administrative segment: 1 ORI number 2 Incident number 3 Incident date/hour 4 Exceptional clearance indicator 5 Exceptional clearance date Offense segment: 6 UCR offense code 7 Attempted/completed code 8 Alcohol/drug use by offender 9 Type of location 10 Number of premises entered 11 Method of entry 12 Type of criminal activity 13 Type of weapon/force used 14 Bias crime code Property segment: 15 Type of property loss 16 Property description 17 Property value 18 Recovery date 19 Number of stolen motor vehicles 20 Number of recovered motor vehicles 21 Suspected drug type 22 Estimated drug quantity 23 Drug measurement unit Victim segment: 24 Victim number 25 Victim UCR offense code 26 Type of victim 27 Age of victim 28 Sex of victim 29 Race of victim 30 Ethnicity of victim 31 Resident status of victim 32 Homicide/assault circumstances 33 Justifiable homicide circumstances 34 Type of injury 35 Related offender number 36 Relationship of victim to offender Offender segment: 37 Offender number 38 Age of offender 39 Sex of offender 40 Race of offender Arrestee segment: 41 Arrestee number 42 Transaction number 43 Arrest date 44 Type of arrest 45 Multiple clearance indicator 46 UCR arrest offense code 47 Arrestee armed indicator 48 Age of arrestee 49 Sex of arrestee 50 Race of arrestee 51 Ethnicity of arrestee 52 Resident status of arrestee 53 Disposition of arrestee under 18 ----------------------------------- Because assault is integral to all violent offenses, it is unaffected by the hierarchy rule in single-victim incidents. Moreover, the definition of aggravated assault is essentially the same for both reporting systems. The difference in treatment of aggravated assault by the two systems lies in classification and recording. Summary UCR collects assaults in five categories: firearm; knife or cutting instrument; other dangerous weapon; hands, fists, feet, and so on; and other simple, not aggravated. Though presence of a weapon indicates aggravated assault, if an incident results in only minor injuries and no information about the weapon is recorded, then it may be misclassified as a simple assault. NIBRS restructures assault into three categories: aggravated, simple, and intimidation. The collection of these characteristics by incident and a broader collection of weapon information reduce the possibility of misclassification. Expanding the hotel rule for NIBRS had more impact on crime statistics. The hotel rule applies to "[b]urglaries of hotels, motels, lodging houses, and other places where lodging of transients is the main purpose . . . This principle of scoring dictates that if a number of dwelling units under a single manager are burglarized and the offenses are most likely to be reported to the police by the manager rather than the individual tenants, the burglary should be scored as one offense.***Footnote"3: FBI, UCR Hnadbook, 1984, p.20.*** NIBRS expands the hotel rule to include temporary rental storage facilities, such as "mini-storage" and "self-storage" buildings.***Footnote 4: FBI, UCR Handbook, NIBRS ed., 1992, p. 13.*** If some mini- storage lockers under the same management are broken into, Summary UCR would count each as a separate burglary whereas NIBRS could count them as one. The lower burglary rates in NIBRS indicate the frequency of this type of break-in. Distributions For total Index crime, the distribution of rate changes shows that 92.5% of the cases have rate difference between -0.5% and 15.5%. Burglary rates that were lower in NIBRS than in Summary UCR could influence total property and total Index crime rates to the degree that they also could have negative values. Almost 48% of the cases differed from -0.5% to +0.5% in crime rates, comparing NIBRS to Summary UCR; 39% of all cases had no change. Cases whose rate was lower when comparing NIBRS counts to Summary UCR counts comprised 4.9% of all cases studied. Overall, 96.4% of all cases had an index crime rate difference of +15.5% or less. For violent crime, the distribution of rate differences show that 98.9% of cases have differences between 0 and 15.5% for total violent crime. Cases with a difference from 0 to 0.5% were 84.3% of the total; 82.2% of all cases had no difference. For total property crime, rates were -0.5% to +15.5% higher for 91.7%. In 48% of the cases the difference between the rates resulting from the two reporting systems was -0.5% to 0.5%, 40.1% of all cases having no difference. Cases whose rate was lower when comparing NIBRS counts to Summary UCR counts comprised 5.1% of all cases studied. Overall, a total of 95.8% of cases showed a property crime rate difference of +15.5% or less. Jurisdictions with a population greater than 25,000 The largest jurisdictions in the dataset (1991-96) covered populations between 25,000 and 250,000. The overall results of the 378 cases from these larger jurisdictions are very similar to the results from all jurisdictions studied (the original 4068 cases). The Index crime rate for NIBRS was about 2% higher than for Summary UCR. The NIBRS violent crime rate was higher by less than 1%, and the NIBRS property crime rate was higher by slightly more than 2%, on average. Murder rates from NIBRS and Summary UCR were the same. The average difference in rates for rape from the two systems was slightly more than 1%. Robbery and aggravated assault rates were higher by an average of less than 1%, going from Summary UCR to NIBRS. NIBRS burglary rates were lower by an average of slightly more than 1%. NIBRS larceny rates were higher than those from Summary UCR by slightly over 3%, and NIBRS motor vehicle theft rates, higher by just over 4%, on average. The distribution of the rate differences for large jurisdictions was also similar to the results of all jurisdictions studied. Among cases from large jurisdictions, 92.3% had a difference of 5.5% or less (including any lower rates) in the total index crime rate, as measured by Summary UCR and NIBRS. Among cases from jurisdictions with nonzero populations, 90.0% had a negligible difference. The distribution within this range was spread out. For example, 31.5% of these cases experienced an Index crime rate difference from -0.5% to 0.5%. A few large jurisdictions (6.3%) showed no difference, and in 9.8% of the cases the Index crime rate was lower using NIBRS data. The proportion of cases with a violent crime rate change of 5.5% or less is 98.4% for large jurisdictions. More than half of the large jurisdictions had a higher rate in 0.5% or less of cases. Nearly 40% experienced no change in the violent crime rate when using NIBRS rather than Summary UCR. Among cases from populations 25,000 or over, 93.1% experienced a property crime rate change of 5.5% or less, compared to 89.6% of all cases with nonzero populations. The property crime rate did not differ for nearly 8% and was lower for more than 10% of the large jurisdictions. Although a smaller proportion of large jurisdictions had no change at all, and a higher proportion showed lower rates compared to all nonzero populations, similar proportions had a 5.5% or less difference in rates, comparing results from NIBRS and Summary UCR. The same pattern holds true for proportions of jurisdictions with a 15.5% or less change in rates. Results for specific crime categories, 1996 Rates for larceny and motor vehicle theft were found to be substantially higher and that for aggravated assault, a slightly higher, comparing NIBRS to Summary UCR. The 1996 NIBRS incident file was analyzed to see what characteristics coincided with these higher rates. Incidents with more than one victim and/or more than one offender were excluded from this analysis. Larceny There were 381,233 incidents in the study data with at least one count of larceny in NIBRS in 1996. These can be broken into two categories: 369,690 single-offense incidents and 11,543 multiple-offense incidents. Of the multiple-offense incidents, 9,601 had a combination of 1 or 2 larcenies and at least 1 other Index crime. The remaining 1,942 multiple-offense incidents involved 2 or 3 counts of larceny and no other Index crime. Under the hierarchy rule of the Summary UCR, the incidents involving only larceny would each be counted as one incident, regardless of whether there were multiple larcenies. Because larceny is at the bottom of the hierarchy, the 9,601 multiple-offense incidents involving larceny and another Index crime would be counted under the concurrent Index crime, not the larceny. From the above figures for 1996, 371,632 incidents including larceny were counted under Summary UCR. Under the NIBRS system the number of incidents involving larceny increased 2.6%. This higher rate can be attributed to the hierarchy rule. The total counts of larceny under NIBRS would be 383,256 (the number of incidents involving 1 larceny, plus twice the number of incidents involving 2 larcenies, plus 3 times the number of incidents involving 3 larcenies, regardless of what other Index crimes may have been concurrent). Under Summary UCR, due to the hierarchy rule, the number of counts of larceny equals the number of incidents that involve only larceny. There was a 3.1% difference between NIBRS larceny counts and Summary UCR larceny counts. This 0.5% higher level over the number of incidents was due to the reporting of multiple counts of larceny within an incident. Three of the nine States with NIBRS data had a disproportionate rise in larceny rates. The three States together accounted for 17% of all larcenies, but for 80% of the burglary- larceny combinations and for 65% of the larceny- larceny combinations. While the average percent difference in larceny rates for all nine States was 3.4%, the average for the three States was 19.1% (not shown in a table). In multiple-offense incidents some records from the three States included larceny segments contrary to FBI instructions. For example, the FBI defines burglary as breaking and entering with intent to commit a felony or theft. Theft is included in the definition of burglary and should not be reported as a concurrent larceny.***Footnote 5: FBI, Uniform Crime Reporting Handbook, NIBRS edition, 1992, p. 13.*** Index crime combinations Number of for 1 victim, 1 offender incidents Total 2,106 Violent Murder 5 Rape 5 Robbery 157 Assault 41 Murder, robbery 2 Rape, robbery 1 Violent and property Rape, burglary 1 Robbery, aggravated assault 3 Robbery, burglary 5 Robbery, larceny 1 Assault, larceny 1 Robbery, aggravated assault, burglary 1 Property Burglary 723 Larceny 1,101 Burglary, larceny 44 Larceny, larceny 15 Aggravated assault Index crime combinations Number of for 1 victim, 1 offender incidents Total 725 Violent Murder 3 Rape 58 Robbery 75 Rape, robbery 2 Violent and property Rape, burglary 7 Robbery, burglary 2 Robbery, motor vehicle theft 3 Robbery, larceny 1 Robbery, burglary, motor vehicle theft 1 Property Burglary 294 Motor vehicle theft 41 Larceny 225 Burglary, larceny 10 Motor vehicle theft, larceny 3 Combination of assaults that occurred in incidents with one or more assaults Number of for 1 victim, 1 offender incidents Total 185,408 Aggravated 35,452 Simple 118,451 Intimidation 30,454 Aggravated, simple 393 Aggravated, intimidation 222 Aggravated, simple, intimidation 43 Simple, intimidation 393 In the 1996 data, more than 8,200 incidents were recorded as burglary-larceny combinations. Though a small number of these may have included a count of larceny distinctly separate from the burglary, the majority of these should have been recorded as burglary only. Combination of assaults that Number occurred in incidents with of larceny for 1 victim, 1 offender incidents Total 1,072 Aggravated 226 Simple 707 Intimidation 123 Aggravated, simple 11 Aggravated, intimidation 1 Aggravated, simple, 1 intimidation Simple, intimidation 3 These States have penal codes with reporting implications that differ from FBI standards. Some State penal codes define burglary as breaking and entering only and mandate recording larceny if property was stolen. Others mandate recording every crime an arrestee might be charged with. As a consequence, the crime rates produced by NIBRS in these States are erroneously inflated. Motor vehicle theft In the 1996 NIBRS incident file, there were 40,667 incidents involving motor vehicle theft; 38,561 were single- offense incidents. There were 2,106 multiple-offense incidents that involved a single motor vehicle theft and 1 or more other Index crimes: 1,162 involved larceny, 774 involved burglary, and 44 involved all 3 property crimes (burglary, larceny, and motor vehicle theft). Two hundred twenty-three incidents of motor vehicle theft involved one or more violent offenses. A total of 990 incidents involved motor vehicle theft and what Summary UCR would classify as a more serious offense, so that approximately 2.4% of the motor vehicle thefts would have been suppressed by the hierarchy rule. Some agencies ranked larceny above motor vehicle theft in the hierarchy. However, incidents involving both motor vehicle theft and larceny are classified as motor vehicle theft in Summary UCR.***Footnote 7: FBI, Uniform Crime Reporting Handbook, 1984, p. 35.*** These agencies would also show more motor vehicle thefts after correction of this reporting error. Aggravated assault There were 36,110 incidents involving aggravated assault in the data file. Of these, 35,385 were single-offense incidents. The 725 multiple- offense incidents involved 1 count of aggravated assault and at least 1 other Index crime. In 152 multiple-offense incidents an aggravated assault was recorded with at least 1 other violent crime. These cases comprised 0.4% of all aggravated assaults and would be suppressed by the hierarchy rule in Summary UCR reporting. Eighty-four multiple-offense incidents involved both aggravated assault and robbery. Because incidents with only one victim and one offender were selected, and assault is one element of robbery, reporting both appears to be either an error or another indication of the influence of State penal code on classification for UCR data. There were 185,408 incidents involving one or more types of assault, 184,357 with a single count of aggravated assault, simple assault, or intimidation. When multiple- offense cases were selected in the manner described above, 1,051 incidents had some combination of aggravated assault, simple assault, and intimidation. There were 658 incidents of aggravated assault with at least 1 lesser assault, and 393 incidents of simple assault with intimidation. There were 1,072 incidents involving larceny and at least 1 type of assault. Of these, 226 involved an aggravated assault and larceny; 13 involved aggravated assault, at least 1 lesser assault, and larceny; and 833 involved 1 or 2 lesser assaults and larceny. Conclusion Because of the selection criteria and the voluntary nature of UCR reporting, the cases in this analysis do not constitute a random sample. Although one must be careful in drawing conclusions on a national level, there is no reason to believe the results are skewed. The large sample size (1,131 jurisdictions, 4,068 cases) suggests that results may be applicable to similar jurisdictions. To accurately compare NIBRS and Summary UCR data, several factors need to be considered: artifactual differences, non-uniform reporting practices, and reporting errors. Artifactual differences between the two reporting systems include expansion of the hotel rule and modified or restructured crime definitions in NIBRS, and the use of the hierarchy rule in Summary UCR. By design NIBRS measures different aspects of crime than Summary UCR. Where Summary UCR measures the overall number of incidents for an index of crime, NIBRS measures both the overall number of incidents and the occurrence of each type of crime within every incident. NIBRS reporting gives more detail about offenses that occur during the commission of a crime without reporting significantly higher crime rates in most cases. However, NIBRS is more prone to non-uniform reporting practices and programming errors, whereas Summary UCR is more prone to reporting errors. Analysis for this report identified several instances of non-uniform reporting practices. After 1996 the FBI began a series of data edits to help reconcile some of those practices. These edits, which were not in place at the time the data used in this research were collected, identify crime combinations that do not comply with FBI reporting standards. A chart denoting these combinations on page 115 of Uniform Crime Reporting National Incident-Based Reporting System, Volume 4, Error Message Manual is reproduced in . Incident reports that contain any of the indicated combinations (occurring to the same victim) are sent back to the reporting agency for correction. The data edits do not cover the non-uniform reporting practices that had the most impact on the crime statistics in this study. For example, although a combination of burglary and larceny in a single incident is contrary to FBI standards, this combination is not included in the data edits because the FBI makes an allowance for the rare occasion in which this combination may legitimately be reported. This study found this combination in certain jurisdictions because the State or local penal code required it. Consequent to this study the FBI began addressing this issue in its State-level training program. It is important for jurisdictions to collect and maintain data in compliance with their penal codes. Uniformity is also critical for data submitted for reporting at the national level. Many States have worked on developing a software program that can automatically convert the crime codes as defined by State penal codes to the appropriate NIBRS crime code. At least one State has successfully implemented an automated State-to-NIBRS crime code conversion system. Others have been unsuccessful due to State codes overlapping more than one NIBRS code. In these cases, the conversion is done manually while the States develop new programs capable of distinguishing the complex codes. As noted in pages 2 and 8, the reporting errors in Summary UCR identified in this report occur largely from undercounting aggravated assault and misapplying the hierarchy rule for motor vehicle theft. Further study is required to determine how much impact the misapplication of the hierarchy rule has on motor vehicle theft and larceny statistics. The design of NIBRS automatically reduces this type of reporting error. As States and jurisdictions are made aware of these potential errors in Summary UCR reporting, they can be corrected before switching to NIBRS. Another concern not addressed by this analysis is the effect NIBRS reporting might have on the rate of aggravated assault. The effects of misclassifying aggravated assaults as simple assault depend on the proportion of misclassified aggravated assaults. To study such effects requires having the original Summary UCR reported by an agency. BJS had NIBRS data that had been converted to Summary UCR. It is not meaningful to compare Summary UCR data from one year to NIBRS data from another. When the two reporting methods are compared within a given year, external variables are controlled for. When one year of data is compared to another, many more variables, such as the prevalence and reporting of crime, or errors in applying the hierarchy rule are introduced. Not all of these variables can be identified or measured, compounding the error in the results. After several years of NIBRS data have been collected, it would then be more meaningful to compare one year of NIBRS data to another. The FBI has begun publishing a series of topical studies using NIBRS data. The first one, "Juvenile Female Crime: A Special Study" appeared in Crime in the United States, 1997, section V; "Incidents of Family Violence" appeared in the 1998 report. Studies covering various topics will be published in other sections of Crime in the United States in future years. For the more expansive tables, the FBI currently publishes statistics using only Summary UCR data and NIBRS data that have been converted back to Summary UCR. They are preparing a report that compares data submitted in Summary UCR format to data submitted in NIBRS. Plans for a NIBRS publication comparable to Crime in the United States are under development. Methodology For comparisons the rate per 100,000 population was calculated for each crime category in each case. For a positive percent change to indicate that the NIBRS rate is higher than the corresponding Summary UCR rate, the following formula was used. 100 x (NIBRS Rate - Summary UCR Rate) ------------------------------ Summary UCR Rate In a small number of cases for specific offenses the Summary UCR count was zero, and the NIBRS count was greater than zero. The resulting percent difference was mathematically undefined. These cases were included in the denominators in the calculation of percentages of cases. Negative changes, indicating that the NIBRS rate is lower than the corresponding Summary UCR rate, are possible for burglary (and consequently for total property crime and total index crime) due to the expansion of the hotel rule in NIBRS. To clearly see the effect of NIBRS on the crime rates of larger jurisdictions, the same analysis was done using a subset of these 4,068 cases. Jurisdictions with populations over 25,000 and that were not combined reporters were selected. Combined reporters are agencies that report data for at least one other jurisdiction in addition to their own. To ensure this analysis included only jurisdictions with populations over 25,000, combined reporters were excluded. 378 cases were selected. None of the jurisdictions reporting NIBRS data between 1991 and 1996 covered populations over 250,000. The Bureau of Justice Statistics is the statistical agency of the U.S. Department of Justice. Jan M. Chaiken, Ph.D., is director. BJS Special Reports address a specific topic in depth from one or more data sets that cover many topics. Ramona Rantala, BJS Statistician, wrote this report. Thomas Edwards, FBI Systems Analyst, assisted with the assembling of NIBRS data and converting to the Summary UCR format. Victoria Major, Yoshio Akiyama, Sharon Propheter, James Nolan, and Christopher Enourato, all FBI CJIS staff, assisted with reviewing this report. Tom Hester edited and produced the report, assisted by Ellen Goldberg. Jayne Robinson produced the report for final printing. July 2000, NCJ 178890 End of file ih 7/7/00 revised 2/23/01 ih